Sustained Preservation of Cognition and Prevention of Patient-Reported Symptoms with Hippocampal Avoidance during Whole-Brain Radiotherapy for Brain Metastases: Final Results of NRG Oncology CC001

Abstract

PURPOSE:

Initial report of NRG Oncology CC001, a phase III trial of whole-brain radiotherapy plus memantine (WBRT+memantine) with or without hippocampal avoidance (HA), demonstrated neuroprotective effects of HA with median follow-up less than 8 months. Herein, we report the final results with complete cognition and patient-reported outcomes and longer-term follow-up exceeding one year.

METHODS:

Adult patients with brain metastases were randomized to HA-WBRT+memantine or WBRT+memantine. The primary endpoint was time to cognitive function failure, defined as decline using the reliable change index on the Hopkins Verbal Learning Test-Revised (HVLT-R), Controlled Oral Word Association (COWA), and/or the Trail Making Tests (TMT) A and B. Patient-reported symptom burden was assessed using the M.D. Anderson Symptom Inventory with Brain Tumor Module and EQ-5D-5L.

RESULTS:

Between July 2015 and March 2018, 518 patients were randomized. Median follow-up for alive patients was 12.1 months. The addition of HA to WBRT+memantine prevented cognitive failure (adjusted hazard ratio, 0.74, p=0.016), and was associated with less deterioration in TMT-B at 4 months (p=0.012) and HVLT-R Recognition at 4 (p=0.055) and 6 months (p=0.011). Longitudinal modeling of imputed data showed better preservation of all HVLT-R domains (p<0.005). Patients who received HA-WBRT+Memantine reported less symptom burden at 6 (p<0.001 using imputed data) and 12 months (p=0.026 using complete-case data; p<0.001 using imputed data), less symptom interference at 6 (p=0.003 using complete-case data; p=0.0016 using imputed data) and 12 months (p=0.0027 using complete-case data; p=0.0014 using imputed data), and fewer cognitive symptoms over time (p=0.043 using imputed data). Treatment arms did not differ significantly in overall survival, intracranial progression-free survival or toxicity.

CONCLUSION:

With median follow-up exceeding one-year, HA during WBRT+memantine for brain metastases leads to sustained preservation of cognitive function and continued prevention of patient-reported neurologic symptoms, symptom interference and cognitive symptoms with no difference in survival or toxicity.

Conventional WBRT, using lateral opposed fields to encompass the whole-brain parenchyma, has been used for well over half a century but has raised toxicity concerns with modern trials demonstrating cognitive deterioration in the majority of patients treated with conventional WBRT [1] Hippocampal neurogenesis, the generation of new hippocampal neurons from neural stem cells located within the subgranular zone of the hippocampal dentate gyrus, has been shown in preclinical and clinical studies to be important to cognitive function in the adult brain and sensitive to relatively low doses of radiation.[2,3]

Our research group hypothesized that the exquisite radiation sensitivity of hippocampal neurogenesis may contribute to radiotherapy-induced cognitive toxicity and that techniques using intensity-modulated radiotherapy (IMRT) to deliver therapeutic doses of WBRT while limiting radiation dose to the bilateral hippocampal dentate gyri, called hippocampal avoidant WBRT (HA-WBRT), could prevent this WBRT-induced cognitive toxicity.[4,5]

Based on promising results from NRG Oncology’s RTOG 0933, a multi-institution phase II trial of HA-WBRT for patients with brain metastases [6], NRG Oncology launched CC001, a multi-institutional phase III trial comparing HA-WBRT plus memantine to WBRT plus memantine for patients with brain metastases. On preliminary analysis, this trial met its primary endpoint and demonstrated better cognitive and patient-reported outcomes following HA added to WBRT plus memantine with no difference in survival or toxicity[7]. While these initial results established HA-WBRT as a standard of care for brain metastasis patients who plan to receive WBRT[8–10], the preliminary analysis did not include data on all enrolled patients and had a median follow-up for alive patients of 7.9 months.

NRG CC001 pre-specified a final statistical analysis following complete cognitive and patient-reported outcome data collection from all enrolled patients, Herein, we report this final data analysis, with median follow-up for alive patients now exceeding one year, to determine if the benefits of HA during WBRT plus memantine are sustained in longer-term survivors. On exploratory analyses, we also report correlations between cognitive function and baseline factors and patient-reported psychological outcomes.

METHODS

TRIAL PATIENTS

Adult patients (≥18 years of age) with brain metastases outside a 5-mm margin around either hippocampus were eligible. Eligibility criteria included Karnofsky Performance Status (KPS) of ≥70 and pathologically proven diagnosis of solid tumor malignancy. Prior resection of brain metastases or radiosurgery was allowed. Exclusion criteria included radiographic evidence of hydrocephalus or other architectural distortion of the ventricular system, leptomeningeal metastases, planned cytotoxic chemotherapy during WBRT, prior WBRT, allergy to memantine, or current use of other NMDA antagonists. The complete eligibility criteria are provided in the trial protocol (Appendix online only). Institutional review board approval was required. All patients were required to provide informed consent.

TRIAL DESIGN AND TREATMENT

Patients were stratified according to recursive partitioning analysis (RPA) class (I vs. II) and prior therapy (none vs. radiosurgery or surgical resection) and randomly assigned, using a permuted block procedure, to WBRT+memantine or HA-WBRT+memantine.[11] [12]

In both study arms, memantine for twice-daily dosing was prescribed 5-mg morning dose week 1, 5-mg twice a day week 2, morning dose 10-mg and evening dose 5-mg week 3, and 10-mg twice a day weeks 4 through 24.[13] Extended release formulation was prescribed 7-mg daily dose week 1, 14-mg daily dose week 2, 21-mg daily dose week 3, and 28-mg daily dose weeks 4 through 24.

For both study arms, the prescribed WBRT dose was 30 Gy in 10 fractions. Hippocampal contouring and HA-WBRT planning directives have been previously described[4,6] with bilateral hippocampal contours manually generated on the fused thin slice MRI-CT image set and expanded by 5 mm to generate the HA-region. The planning target volume (PTV) was defined as the whole-brain parenchyma excluding the HA-region; no set-up margin was added to the PTV. IMRT was utilized to deliver the conformal radiotherapy plan for patients treated with HA-WBRT (protocol details see appendix online only).

Before enrolling patients on this trial, all participating sites completed a credentialing exercise in which they would generate a HA-WBRT treatment plan on a sample case that would be reviewed centrally. For the first case planned for HA-WBRT by a radiation oncologist, rapid central review of hippocampal contours and HA-WBRT planning was conducted in real-time before initiation of treatment. If the initial plan was deemed acceptable, subsequent treatment plans were reviewed post-treatment in a timely manner to provide close monitoring, assessment of plan quality, and ongoing feedback (Appendix Table A1, online only).

ASSESSMENTS

Before randomization each patient underwent a baseline evaluation consisting of history and physical examination, neurological examination, performance status, and thin-slice MRI, and completed cognitive testing and measures of patient-reported QOL and symptom burden. All baseline evaluations along with assessment of adverse events were repeated at month 2, 4, 6, and 12. The cognitive testing was administered by a trained, certified member of the site study team (see online protocol for details of training). The same validated battery of cognitive tests utilized in the prior memantine trial was administered in this trial to assess learning and memory (Hopkins Verbal Learning Test-Revised [HVLT-R]), verbal fluency (Controlled Oral Word Association [COWA]), processing speed (Trail Making Test Part A [TMT-A]), and executive function (Trail Making Test Part B [TMT-B]).[13–15] Higher HVLT-R and COWA scores indicate better performance (i.e. listing more correct words) while higher TMT-A and TMT-B indicate worse performance (i.e. longer time to complete the task). QOL and symptom burden were assessed by the EQ-5D-5L and the MD Anderson Symptom Inventory with Brain Tumor module (MDASI-BT), respectively.[16,17] The MDASI-BT is a self-report measure of symptom burden and symptom interference. It consists of 22 symptom items rated 0 (not present) to 10 (as bad as you can imagine) and 6 interference items rated 0 (did not interfere) to 10 (interfered completely). Symptom item and interference items can be further categorized into symptom factors, and Cognitive and Neurologic symptom factors were pre-specified items of interest.

For the index and visual analog scale (VAS) score from the EQ-5D-5L, higher scores indicate better QOL, while for the MDASI-BT factors, Symptom Severity and Symptom Interference, higher scores indicate more symptoms. All treatment-related toxicities and adverse events were recorded according to NCI Common Terminology Criteria for Adverse Events (CTCAE) version 4.0.

END POINTS

The primary endpoint was time to cognitive failure, defined as cognitive decline determined by the reliable change index (RCI) on at least one of the cognitive tests.[13] Patients and clinicians were not blinded to treatment assignment, although the Neurocognitive Chair scoring the cognitive tests was blinded to treatment assignment. Secondary endpoints included intracranial progression-free survival (PFS), overall survival (OS), toxicity, patient-reported symptoms (focusing on Symptom Severity, Symptom Interference, Neurologic factor, and Cognitive factor subscale scores as well fatigue, Neurologic factor items, and Cognitive factor items), QOL, and cognitive function as measured separately by the individual cognitive tests and the Clinical Trial Battery Composite score.[13]

STATISTICAL ANALYSIS

The analysis was conducted on all randomized patients using intent-to-treat. The cumulative incidence approach was used to estimate the time to cognitive failure to account for the competing risk of death. Gray’s test was used to test for a statistically significant difference in the distribution of cognitive failure times.[18] OS and intracranial PFS were estimated using the Kaplan-Meier method, and differences between treatment arms were tested using the log rank test.[19,20] OS and intracranial PFS were measured from the date of randomization to the date of intracranial progression for intracranial PFS only, death or the last follow-up date on which the patient was reported alive. Unadjusted Cox proportional hazards models were used to obtain HRs and 95% confidence intervals (CI) for OS and intracranial PFS while cause-specific Cox models were used for time to cognitive failure.[21] Cognitive deterioration for each cognitive test was defined using the RCI. Between-arm comparisons of categorical variables, such as cognitive deterioration and adverse events, were tested using chi-square tests. Change from baseline scores for QOL and symptom items were compared between arms using a t-test. Trends across time on the MDASI-BT and the standardized cognitive tests (adjusted for age, education, and gender) were modeled with mixed effects models using maximum likelihood estimation. Individual MDASI-BT items were compared between arms for Cognitive and Neurologic factors, if the factor score was significantly different between arms, and for fatigue. Tests at 6 and 12 months were conducted within the framework of the longitudinal model to account for data considered missing at random and allow for adjustment from covariates of interest. These tests were two-sided using a significance level of 0.05 for the MDASI-BT and cognitive tests with Hochberg’s procedure used to adjust for multiplicity for the multiple MDASI-BT domains analyzed.[22] As a sensitivity analysis due to missing data, models were run using imputed data for alive patients with consent who were missing data with tests conducted at 6 and 12 months also performed. Multiple imputation employed a Markov chain Monte Carlo (MCMC) method with 20 iterations.

As an exploratory analysis, patient-reported psychologic distress at each time point was correlated with baseline cognitive function using Spearman correlation coefficients. Correlation coefficients of clinical interest are with at least moderate strength, defined as ρ>0.20.[23,24] The symptom burden items of interest are the “distressed (upset),” “sad,” and “mood” items; the EQ-5D-5L item of interest is the “depression/anxiety” item; and, the MDASI-BT Cognitive Factor was also examined. Spearman correlation coefficients were also used to assess possible correlations between baseline neurocognitive function and RPA Class and disease-specific graded prognostic assessment (DS-GPA) score. Standardized cognitive test scores were used in the correlation analyses.

RESULTS

STUDY PATIENTS

Between July 13, 2015, and Mar 12, 2018, 518 patients were randomly assigned to WBRT+Memantine or HA-WBRT+Memantine from 112 participating institutions in the USA and Canada (Figure 1). Sixty-two patients were found to be ineligible, 32 in the WBRT+Memantine arm and 30 in the HA-WBRT+Memantine arm, most commonly due to imaging not completed per protocol. Baseline characteristics, including baseline cognitive function, were well balanced between the study arms and have been previously reported [7]. Median age was 61.5 years (20–91) and most patients had primary lung cancer (57.7%). The median follow-up for alive patients was 12.1 months (range: 0.03–15.6). Compliance rates for cognitive testing and patient reported outcomes are provided in Appendix Table A1, online only.

Figure 1.

CONSORT diagram.

HA=Hippocampal Avoidance; WBRT=Whole Brain Radiation Therapy; HVLT-R=Hopkins Verbal Learning Test-Revised.

TREATMENT OUTCOMES

Primary Analysis and Cognitive Outcomes

The addition of HA to WBRT+memantine significantly reduced cognitive failure risk (unadjusted HR=0.76, 95%CI: 0.60–0.98, p=0.029) (Figure 2, Table A2). In the adjusted cause-specific analysis, the treatment effect was more pronounced in favor of HA-WBRT+Memantine (HR=0.74, 95%CI: 0.58–0.95, p=0.016) (Table 1). Age ≤61 years also predicted for lower cognitive failure risk (HR=0.61, 95% CI: 0.47–0.80, p=0.0003). The addition of an interaction term between age and treatment arm did not individually change the significance of age and treatment arm and was itself not significant, indicating that the effect of HA was independent of age.

Figure 2.

Time to cognitive failure. WBRT, whole-brain radiotherapy; HA-WBRT, hippocampal avoidant whole-brain radiotherapy.

Table 1.

Time to Cognitive Failure Multivariate Cox Proportional Hazards Model

Variables	Hazard Ratio	95%CI	p-value
Age (<=61 vs. >61[RL])	0.642	0.485–0.850	0.0020
RPA Class (I vs. II[RL])	1.198	0.815–1.760	0.36
Prior Radiosurgery(No vs. Yes[RL])	0.817	0.619–1.078	0.15
Prior Surgical Resection (No vs. Yes[RL])	1.128	0.857–1.484	0.39
Metastasis (Brain Only vs. Brain and Other Sites[RL])	1.188	0.880–1.602	0.26
Treatment Arm (HA- WBRT+Memantine vs. WBRT+Memantine [RL])	0.739	0.577–0.945	0.016*

^*There was no significant effect of an interaction between age and treatment arm (i.e. effect of treatment does not differ by age). Patients lost to cognitive failure/total: 261/518; patients lost to death: 105.

RL, reference level; RPA, recursive partitioning analysis; HA-WBRT, hippocampal avoidance whole brain radiotherapy; WBRT, whole brain radiotherapy

Analysis of each cognitive test separately revealed that the HA-WBRT+Memantine arm was less likely to have deterioration in TMT-B (23.3% vs. 40.4%, respectively, p=0.012; Table A3b) and HVLT-R Delayed Recognition at 4 months (14.0% vs. 24.8%, p=0.05) and at 6 months (17.6% vs. 36.3%, p=0.011; Table A3c). There were no cognitive test-specific inter-arm differences in deterioration status at 2 months (Table A3a) and 12 months (Table A3d). Trends over time are presented in Figures 3 and A1. Mixed effects modeling using complete-case data revealed estimates in favor the HA-WBRT+memantine arm with no significant treatment arm differences (Table A4a), but mixed effects modeling using imputed data demonstrated better preservation over time of HVLT-R Total Recall (p=0.0006), Delayed Recognition (p=0.043), and CTB Composite score (p=0.027) scores in the HA-WBRT+Memantine arm (Table A4b).

Figure 3.

Hopkins Verbal Learning Test-Revised (HVLT-R) (A) Total Recall, (B) Delayed Recall and (C) Recognition. Higher score indicates better performance. WBRT, whole-brain radiotherapy; HA-WBRT, hippocampal avoidant whole-brain radiotherapy.

Patient-Reported Outcomes

Using complete-case data, patients on the HA-WBRT+Memantine arm experienced less Symptom Interference at 6 months and 12 months (estimate=−5.07, standard deviation [SD]=1.69 p=0.003 and estimate=−10.16, SD=3.36 p=0.0027, respectively) and less Symptom Burden at 12 months (estimate=−0.45, SD=0.20, p=0.026) (Table 2). Cognitive factor and Neurologic factor did not show any significant treatment effects using complete-case data (Table A5a). After applying Hochberg’s multiplicity adjustment on imputed data, patients on the HA-WBRT+Memantine arm experienced less Symptom Burden and less Symptom Interference at 6 months (estimate=−1.37, SD=0.33, p<0.0001 and estimate=−1.93, SD=0.61, p=0.0016, respectively) and 12 months (estimate=−2.60, SD=0.63, p<0.0001 and estimate=−3.80, SD=1.17, p=0.0014, respectively; Table 2). At 6 months, patients on the HA-WBRT+Memantine arm reported a smaller change from baseline in difficulty remembering things (mean=0.21, SD=3.21 vs. mean=1.26, SD=1.78, p=0.016).

Table 2.

Test of Treatment Arm Difference at 6 and 12 Months

	A. 6 Months			B. 12 Months
Variable	Estimate (SD)	p-value	n	Estimate (SD)	p-value	n
Complete Data
Symptom Burden	−0.26 (0.15)	0.083	306	−0.45 (0.20)	0.026	306
Symptom Interference	−5.07 (1.69)	0.003	305	−10.16 (3.36)	0.0027	305
Cognitive factor	−0.05 (0.18)	0.77	306	0.09 (0.25)	0.72	306
Neurologic factor	0.21 (0.21)	0.32	306	0.40 (0.28)	0.15	306
Imputed Data *
Symptom Burden	−1.37 (0.33)	<0.001	366	−2.60 (0.63)	<0.001	366
Symptom Interference	−1.93 (0.61)	0.0016	366	−3.80 (1.17)	0.0014	366
Cognitive factor	−0.17 (0.18)	0.35	366	−0.05 (0.26)	0.86	366
Neurologic factor	0.13 (0.22)	0.56	366	0.33 (0.31)	0.29	366

t-Test was conducted within the longitudinal model framework. Using Hochberg’s procedure for interference, cognitive factor and neurologic factor (since symptom burden was the primary PRO endpoint powered with a type I error=0.05), the significance levels for the smallest to largest p-values are 0.0167, 0.033, and 0.05.

^*There were 366 patients used in 1 imputation with 20 imputations per model.

Mixed effects modeling using imputed data revealed that patients on the HA-WBRT+Memantine arm experienced fewer cognitive symptoms over time (Cognitive Factor, mean between arm difference=−0.29, p=0.0425; Figure 4 and Table A5b). No differences were seen between study arms at baseline or overtime for the EQ-5D-5L (Table A6).

Figure 4.

Cognitive Factor from M.D. Anderson Symptom Inventory (MDASI) with Brain Tumor module. Higher score indicates more symptoms. WBRT, whole-brain radiotherapy; HA-WBRT, hippocampal avoidant whole-brain radiotherapy.

Survival, Intracranial Progression, and Toxicity

There was no difference between arms in terms of OS (median of 6.3 months HA-WBRT+ memantine vs. 7.6 months, unadjusted HR=1.14, 95% CI: 0.91–1.43, p=0.24; Figure A2) or intracranial PFS (median of 5.0 months HA-WBRT+memantine vs. 5.3 months, unadjusted HR=1.19, 95% CI: 0.97–1.46, p=0.087; Figure A3). Relapses in the HA-region were 11 in the HA-WBRT+ memantine arm vs. 17 in the WBRT+Memantine arm. There was no difference in grade 3+ toxicity without regard to attribution (62.1% vs. 58.7%, respectively, p=0.47) or related to treatment (19.8% vs. 19.3%, respectively, p=0.87; Tables A7–A8) between the WBRT+Memantine and HA-WBRT+Memantine arms, respectively.

Correlation between Baseline Cognitive Function and Patient-Reported Psychologic Distress

Increased patient-reported anxiety/depression at 6 months was associated with lower baseline scores in HVLT-R Total Recall (ρ=−0.26, p=0.0013), Delayed Recall (ρ=−0.30, p<0.001), and Delayed Recognition (ρ=−0.30, p<0.001; Table A9). Increased patient-reported anxiety/depression at 6 months was also associated with lower baseline COWA scores (ρ=−0.20, p=0.013), poorer baseline performance on TMT-B (ρ=−0.27, p=0.001; Table A9) and worse CTB composite score (ρ=−0.30, p=<0.001; Table A9). Improved mood at 6 months was associated with lower baseline scores in HVLT-R Delayed Recall (ρ=−0.21, p=0.01). Increased feelings of being distressed (upset) at 12 months was associated with poorer baseline performance on TMT-B (ρ=−0.20, p=0.042; Table A10).

There were no differences in any of the standardized cognitive test scores between patients in RPA Class I vs. II (Table A11). Similarly, there was no correlation between DS-GPA and baseline standardized cognitive function (Table A12).

DISCUSSION

With median follow-up exceeding 12 months and with complete cognitive and patient-reported outcomes data now collected, this phase 3 trial demonstrates sustained benefits of HA during WBRT+Memantine for patients with brain metastases in preserving cognitive function and preventing patient-reported neurologic symptoms, symptom interference and cognitive symptoms. There were no significant differences between treatment arms in toxicity, intracranial progression-free survival, or overall survival. These findings with longer-term follow-up support HA-WBRT plus memantine as a standard of care for good performance status brain metastasis patients planning to receive WBRT.[8–10] These findings impact the estimated 200,000 patients per year who receive WBRT in the United States alone and also potentially the design of radiotherapy for primary brain tumors, although further clinical studies in these patient populations are needed.[3,25]

The difference in absolute rates of cognitive function failure on the WBRT+Memantine arms of NRG/RTOG 0614 (54% at 6 months) [13] and NRG CC001 (69% at 6 months) highlight the potential differences in patients enrolled on NRG/RTOG 0614, conducted during an era when WBRT was more widely employed for limited brain metastases, versus those enrolled on NRG CC001, conducted during an era when radiosurgery was more widely employed for limited brain metastases. This difference also similarly challenges comparisons of absolute rates of cognitive function failure with prior trials of radiosurgery versus WBRT for limited brain metastases. Comparison of cognitive protection strategies should instead focus on the relative-risk reduction derived from the adjusted hazard ratio and reflective of cognitive protection throughout a patient’s survivorship. On NRG/RTOG 0614, memantine provided a 22% relative reduction in cognitive toxicity [13]; on NRG CC001, hippocampal avoidance added to memantine provided an additional 26% relative reduction in cognitive toxicity. Aforementioned results from longer-term follow-up of alive patients on NRG CC001 demonstrate no change in this relative risk reduction with longer median follow-up and emphasize the longer-term cognitive protection offered by HA and memantine during WBRT.

Prevention and/or improvement of neurologic signs and symptoms are integral to brain metastasis management, given the potential for uncontrolled brain metastases to cause symptoms and impaired function. With longer follow-up, HA added to WBRT+Memantine demonstrates continued benefits in patient-reported symptoms not just at 6 months, as reported previously [7], but also at 12 months follow-up. These patient-reported benefits at 12 months included fewer neurologic symptoms (Symptom Burden) less interference of neurologic symptoms in daily life (Symptom Interference) and were observed using both complete-case and imputed data. In addition, complementing the continued benefits in tested cognitive function and supporting the cognition-specific rationale for hippocampal avoidance, patients on the HA-WBRT+memantine arm experienced fewer cognitive symptoms over time (observed using imputed case data) and reported less difficulty remembering (observed using complete-case data) at 6 months. Interestingly, while between-arm differences were noted with MDASI-BT, no such differences were observed with EQ5D-5L. This disparity could be explained by the greater sensitivity of MDASI-BT for clinically relevant neurologic symptoms, as compared to EQ5D-5L developed to estimate general health status in cost-utility analyses.[26] Overall, these patient-reported symptom findings are relatively novel, having never been previously observed in brain metastasis trials, and emphasize not just the importance of including patient-reported outcomes in such trials, but also the importance of HA-WBRT+memantine in achieving the palliative objectives of brain metastasis management.

As exploratory findings, poorer cognitive performance at baseline (i.e., prior to study treatment) was associated with greater subsequent patient-reported psychological distress, specifically anxiety and depression and feelings of being distressed (upset). The strongest correlations (ρ of 0.26–0.30) occurred between poorer learning and memory and executive function at baseline, and greater patient-reported anxiety and depression at 6 months. The link between impaired cognition and subsequent depression and anxiety has been reported in prior studies, especially in older adults [27], but to our knowledge this is the first such link reported in cancer patients. Reasons for these correlations remain unclear and difficult to fully assess using the data collected on NRG Oncology CC001. However, one possible hypothesis that could be evaluated in future studies is the importance of cognition to patient-reported psychological distress. In gerontology, some have hypothesized that depressive symptoms may develop or become exacerbated after the development of cognitive impairment, as a consequence of perceived loss of control over aspects of an individual’s life [28], perceived threats to an individual’s ability to continue living independently [29], and/or impairment in capacity to regulate mood, mobilize resources for social support and participate in activities that promote coping and prevent symptoms of depression [30].

The findings of this trial support the use of HA-WBRT+memantine for the management of brain metastases. However, the precise clinical context in which this approach would be optimal remains an area of ongoing investigation. CCTC CE.7 (ClinicalTrials.gov NCT03550391) is an ongoing phase III trial of radiosurgery versus HA-WBRT+memantine for 5–15 newly diagnosed brain metastases. NRG BN009 is a phase III trial of salvage radiosurgery versus salvage HA-WBRT+memantine for recurrent brain metastases after upfront radiosurgery in patients with high brain metastasis velocity, which has been shown to predict for inferior survival and increased risk of neurologic death [31]. NRG CC009 is a phase III trial of radiosurgery versus HA-WBRT+memantine for small cell lung cancer brain metastases, which have been excluded from prior brain metastasis trials. The summation of these ongoing trials will provide a contemporaneous comparison between radiotherapy approaches for optimal brain metastasis management.

One of the limitations of this study is the inability to blind trial participants and treatment providers to treatment. This is a fairly common limitation to trials comparing different forms of radiotherapy. From the treatment provider perspective, the different radiotherapy approaches require variable approaches to treatment planning, quality assurance and delivery, sometimes requiring the use of different machines or overall treatment times. From the trial participant perspective, blinding to treatment assignment might be logistically difficult and raises ethical issues with respect to the treatment team engaging in active deception [32] However, while lack of treatment assignment blinding could potentially confound patient-reported outcomes, it is not expected to have any meaningful impact on the objective cognitive assessment battery results [33].

Missing data, which by 12 months reached 34% of alive patients who did not withdraw consent, represent an additional limitation to data analysis. This study’s approach to managing this limitation using multiple imputation statistical modeling has been described elsewhere [34], and this report summarizes analyses from both complete-case and imputed data, so that interpretation of modeling can be done in conjunction between both data sets. As an example, the complete-case data analyses showed trends toward significance for the benefits of HA in HVLT-Total Recall and HVLT-Delayed Recognition, but the imputed data analyses for these cognitive tests were statistically significant (Table A4). Similarly, the complete-case data analysis showed trends toward significance for the benefit of HA in Symptom Burden at 6 months, but the imputed data analyses for this MDASI-BT item was statistically significant (Table 2). Notably, at 12 months, analyses of both complete-case and imputed data demonstrated statistical significance for the benefit of HA in Symptom Burden. Assuming data were missing at random, these observations reflect the capacity of the imputation method to correct the bias from missing data and increase the power of statistical analysis since more cases are included. In these sensitivity analyses, biases caused by data that are not missing at random were not addressed.

In conclusion, with longer-term follow-up now exceeding a median of one-year, hippocampal avoidance during WBRT with memantine provides sustained benefits in cognitive function preservation and patient-reported symptom prevention, with no difference observed in toxicity, intracranial progression-free survival, or overall survival compared to standard WBRT and memantine. The correlation between poorer baseline cognitive performance and subsequently increased patient-reported psychological distress, may reflect the multi-dimensional importance of cognitive preservation strategies in the brain metastasis population.

Data Sharing Statement:

All data will be made available per the NCTN Data Archive rules. The link for the archive is: https://nctn-data-archive.nci.nih.gov/